Higher albumin-to-glucose ratio linked with lower 30-day mortality in retrospective study of steroid-treated pneumonia

In a retrospective multicenter analysis of 473 glucocorticoid-treated patients with pneumonia, a higher albumin-to-glucose ratio (AGR) was independently associated with lower 30-day mortality, according to the study abstract. The investigators analyzed data obtained from the Dryad Digital Repository for secondary analysis, assessing the ratio’s ability to help stratify mortality risk among immunocompromised patients receiving glucocorticoids.

The analysis used multivariable Cox proportional hazards regression, receiver operating characteristic curves, mediation analysis, K-means clustering and machine-learning methods. In the multivariable model, each increase in AGR was associated with a lower mortality risk, with an adjusted hazard ratio of 0.82 (95% confidence interval, 0.73–0.91; P < 0.001). The reported optimal discrimination threshold was an AGR of at least 5.0, where the area under the curve was 0.713. Mediation analysis indicated that platelet count accounted for only 9.24% of the association between AGR and mortality.

Clustering based on AGR and platelet count identified three phenotypes with graded mortality risks. In the crude model, the cluster characterized by low AGR and low platelet count had a 4.04-fold higher mortality risk than the cluster with moderate AGR and high platelet count (hazard ratio, 4.04; 95% confidence interval, 2.32–7.03; P < 0.001). After full adjustment, this difference was attenuated and was no longer statistically significant (hazard ratio, 2.02; 95% confidence interval, 0.92–4.42; P = 0.079). Boruta feature-selection analysis identified AGR as a significant predictor. Among six machine-learning models, the ExtraTrees model had the highest reported area under the curve for predicting 30-day mortality, at 0.879, comparable with the other models in the analysis.

The findings are limited by the study’s retrospective design and secondary use of repository data. The authors noted that the clustering and machine-learning analyses lacked internal cross-validation, hyperparameter optimization and external validation, so those results should be considered exploratory and hypothesis-generating. The abstract also reports that AGR’s discriminative performance was comparable to several conventional severity scores and biomarkers in the dataset. External validation is required before the ratio can be applied in clinical practice.

Subject of Research: Albumin-to-glucose ratio and 30-day mortality in immunosuppressed patients with pneumonia receiving glucocorticoids

Article Title: The albumin-to-glucose ratio as an independent predictor of 30-day mortality in immunosuppressed patients with pneumonia receiving glucocorticoids: a retrospective multicenter study

Article References: Yuan, Y., Xing, Z., Zhang, S., & Wang, J. (2026). The albumin-to-glucose ratio as an independent predictor of 30-day mortality in immunosuppressed patients with pneumonia receiving glucocorticoids: a retrospective multicenter study. BMC Infectious Diseases. https://doi.org/10.1186/s12879-026-14437-6

Image Credits: AI Generated

DOI: 10.1186/s12879-026-14437-6

Keywords: albumin-to-glucose ratio, 30-day mortality, pneumonia, immunosuppressed patients, glucocorticoids, risk stratification, biomarkers, retrospective multicenter study, machine learning, platelet count, albumin-to-glucose, ratio